It is demonstrated how information about the undisturbed wave at all frequencies of interest can be retrieved by constrained deconvolution of the outputs of several sensors with different known linear response characteristics, placed on or around a floating structure. A model of the sea state which retains information about the amplitude, phase and direction can then be used to calculate the resulting non-'linear drift forces by means of known quadratic transfer functions. It is shown how, with certain limitations, the wave and drift force calculation can be implemented in real time using frequency domain techniques. The resulting force time history could be used as feed forward information by the controller of a dynamic positioning system, to improve performance and efficiency. This could be further enhanced by applying linear prediction techniques (eg an auto-regressive model) to the force time history.
The dynamic position control problem arises because wave frequency forces are too large to be counteracted, and varying the demanded thrust at these frequencies would cause needless wear and tear on the thrusters. Simple low pass filtering would introduce a lag that would tend to degrade the performance of the system. The most widely used technique to overcome this is Kalman filtering, which was presented in the context of dynamic positioning by Balchen, Jensen and Saelid (1976). The method uses a state space model of the first order dynamics of the vessel to separate out the second order responses and predict" them for one or more steps ahead. What is more desirable however is to estimate the instantaneous forcing so that the appropriate preventative action can be taken before the vessel starts moving off station. He showed that this related to one component of the mean drift force and it was used successfully to improve position keeping.